How Can We Better Inform Our Patients about Post-Heart Transplantation Survival? A Conditional Survival Analysis

Abstract

Background:

Conditional survival (CS) is a dynamic method of survival analysis that provides an estimate of how an individual’s future survival probability changes based on time post-transplant, individual characteristics, and post-transplant events. This study sought to provide post-transplant CS probabilities for heart transplant recipients based on different prognostic variables and provide a discussion tool for the providers and the patients.

Methods:

Adult heart transplant recipients from January 1^st, 2004 through October 18^th, 2018 were identified in the UNOS registry. CS probabilities were calculated using data from Kaplan-Meier survival estimates.

Results:

CS probability exceeded actuarial survival probability at all times post-transplant. Women had similar short-term, but greater long-term CS than men at all times post-transplant (10-year CS 1.8%−11.5% greater [95% CI 1.2%−12.9%]). Patients with ECMO or a surgical BiVAD had decreased survival at the time of transplant, but their CS was indistinguishable from all others by one-year post-transplant. Rejection and infection requiring hospitalization during the first-year were associated with a persistently decreased CS probability.

Conclusions:

In this study, we report differential conditional survival outcomes based on time, patient characteristics, and clinical events post-transplant, providing a dynamic assessment of survival. The survival probabilities, will better inform patients and clinicians of future outcomes.

Introduction:

Heart failure affects more than 6 million Americans, of which an estimated 250,000 have Stage D or end stage heart failure.^1,2 Heart transplantation remains the gold standard for treatment of Stage D heart failure, and there has been a significant increase in the number of adult transplants performed in the United States: 1,926 in 2000 to 3,044 in 2019.³ The increase in heart transplantation has been associated with an increase in post-transplant survival, from a median of 10.5 years to almost 14 years presently.⁴ These increases have been predominantly attributable to improved care in the first year of transplant, where risk of infection and rejection is the greatest. In the United States, heart transplant outcomes are publicly reported and post-transplant survival statistics serve a number of practical purposes. First, benchmark post-transplant survival statistics (1-year and 3-year survival) are used to compare United States transplant programs.⁵ Arguably more importantly, post-transplant survival data provides patients and physicians an estimate of future survival. Numerous prognostic characteristics (donor, recipient, surgical) have been identified to influence overall post-transplant survival. However, the majority of analyses estimate survival in a static manner, while the probability of survival for an individual may change over time. Conditional survival (CS) probabilities (the probability that a patient will survive an additional period contingent on survival to a specified time point) provide estimates of how an individual’s prognosis changes over time. CS probabilities have been frequently reported in oncologic studies and are useful to assess for non-proportional risks over time. Yet, CS probabilities have rarely been applied to solid organ transplantation. This study sought to provide post-transplant CS probabilities for heart transplant recipients based on different prognostic variables and provide a discussion tool for the providers and the patients.

Methods:

This retrospective cohort study utilized the United Network for Organ Sharing (UNOS) registry to identify all adult (age>18 years), primary heart transplant recipients (multi-organ transplant recipients were excluded) from January 1^st, 2004 through October 18^th 2018. Follow-up data were available through June 6^th, 2019. Pre-transplant characteristics collected included demographic (age, sex, heart failure etiology, race, blood type), clinical (renal function, body mass index [BMI], diabetes, cerebrovascular disease, prior sternotomy, prior malignancy, UNOS listing status, patient pre-transplant location, Model for End-stage Liver Disease excluding INR (MELD-XI) score, serum albumin, and pulmonary vascular resistance), and mechanical circulatory support (MCS) use (extracorporeal membrane oxygenation [ECMO], intra-aortic balloon pump [IABP], surgical ventricular assist device). Post-transplant events analyzed included center reported rejection during the index hospitalization (categorized as none, rejection treated with additional anti-rejection agent, and rejection not treated with additional anti-rejection agent), center reported rejection during the first year, and hospitalization for infection during the first year.

Conditional survival was defined as the probability of surviving an additional y years conditional on a patient having survived x years: $S\left(y|x\right)=\frac{S(y+x)}{S\left(x\right)}$. For example, if a patient is two years post-transplant, the probability of surviving an additional five years (seven years in total) was calculated by: $S\left(5|2\right)=\frac{S(5+2)}{S\left(2\right)}$. Survival probabilities used for CS calculations were obtained from Kaplan-Meier survival estimates. Confidence intervals of the CS estimates were calculated using a modification of the Greenwood formula as previously described.⁶ The Institutional Review Board of Columbia University Irving Medical Center previously determined analyses using the UNOS registry (de-identified public data set) as exempt. A two-tailed p-value of less than 0.05 was considered significant. Statistical analyses were performed with SAS Version 9.4 (Cary, NC).

Results:

The study cohort included 30,229 patients. The patients were predominantly male (74.8%) and the majority were middle aged (50–64 years, 52.2%), with smaller proportions young (33.6% 18–49 years) and elderly (14.2% ≥65 years, Table 1). A total of 20,253 (67%) were white, 6,227 (20.6%) were black, 2,418 (8%) were Hispanic, and 4.4% were from other races. The predominant etiology for heart failure was non-ischemic cardiomyopathy (54.5%) followed by ischemic cardiomyopathy (37.1%). Most had normal renal function (GFR>60 mL/min/1.73 m², 62.1%) and 35.0% had Stage III chronic kidney disease (CKD). Nearly half (46.8%) had MCS prior to transplantation including 35.5% with a durable left ventricular assist device, 6.8% with an IABP, and 0.8% with ECMO.

Table 1.

Cohort baseline characteristics

	Number of Patients (%)
Male	22,611(74.8)
Age (years)
18–49	10,157 (33.6)
50–64	15,780 (52.2)
65+	4,293 (14.2)
Race/Ethnicity
Black	6,227 (20.6)
Hispanic	2,418 (8.0)
Other	1,330 (4.4)
White	20,253 (67.0)
Etiology
Ischemic	11,215 (37.1)
Non-ischemic	16,475 (54.5)
HCM	725 (2.4)
RCM	937 (3.1)
Congenital	877 (2.9)
Blood Type
A	12,273 (40.6)
B	4,474 (14.8)
O	12,273(39.1)
AB	1,663 (5.5)
Diabetes	8,313 (27.5)
Prior Malignancy	2,328 (7.7)
Renal Function
GFR>60	18,772 (62.1)
GFR 30–60	10,580 (35.0)
GFR<30	877 (2.9)
BMI
Underweight (<18.5)	665 (2.2)
Normal Weight (18.5–25)	10,006 (33.1)
Overweight (25–30)	11,124 (36.8)
WHO Class I Obese (30–35)	6,681 (22.1)
>WHO Class I Obese (>35)	1,753 (5.8)
Symptomatic Cerebrovascular Disease	1,602 (5.3)
UNOS Status
Status 1A	16,535 (54.7)
Status 1B	9,976 (33.0)
Status 2	2,539 (8.4)
MELD-XI >14.6	6,348 (21.0)
Albumin <3.5 mg/dL	11,457 (37.9)
Patient Location
ICU	8,887 (29.4)
Hospitalized, not in ICU	4,897 (16.2)
Not Hospitalized	16,445 (54.4)
PVR
PVR<3	23,851 (78.9)
PVR 3–4.5	4,897 (16.2)
PVR>4.5	2,116 (7.0)
MCS
IABP	2,056 (6.8)
ECMO	242 (0.8)
LVAD	10,731 (35.5)
BiVAD	816 (2.7)
TAH	333 (1.1)

BiVAD, surgical biventricular assist device; BMI, body mass index; ECMO, extracorporeal membrane oxygenation; GFR, glomerular filtration rate; HCM, hypertrophic cardiomyopathy; IABP, intra-aortic balloon pump; ICU, intensive care unit; LVAD, left ventricular assist device; MELD-XI, Model for End-stage Liver Disease excluding INR score; PVR, pulmonary vascular resistance; RCM, restrictive cardiomyopathy; TAH, total artificial heart; UNOS, United Network for Organ Sharing; WHO, world health organization.

Overall Conditional Survival

Overall one-year post-transplant survival was 90.0% and the median post-transplant survival was 12.9 years. CS probabilities were calculated at 1, 3, 5, and 10 years post-transplantation for 1, 3, 5 and 10-year additional survival (Figure 1). One-year CS improved to 96.1% (95% CI 96.1 to 96.2%) for those who survived to 1 year post-transplant, and remained stable at 96.6% (95% CI 96.5 to 96.7%) and 95.9% (95% CI 95.8 to 96.0%) at 3 and 5 years post-transplant, respectively. Mid-term (3 and 5 year) CS also improved at 1 year post-transplant but predictably declined as time post-transplant increased. Ten-year CS improved slightly at 1 year, but decreased thereafter. At each time point, the five-year CS exceeded survival probability at the time of transplant. For instance, if a patient was alive 5 years post-transplant, the probability of surviving an additional 5 years or 10 years overall was 78.2% (95% CI 77.7 to 78.7%, demonstrated by the green point at 5 years in Figure 1) or 17.5% (95% CI 17.2 to 17.8%, p <0.001) greater than surviving 10 years at the time of transplant. Similarly, the 10 year CS probability (15 year overall survival, demonstrated by the brown point at 5 years in Figure 1) was 49.2% (95% CI 46.1 to 52.4%), 24.7% (95% CI 23.8 to 25.6%, p<0.001) greater than the predicted 15 year survival at the time of transplant.

Figure 1.

Conditional survival for all patients following heart transplantation

Conditional Survival by Sex

Women had improved overall post-transplant survival (median 13.8 years) compared with men (median 12.6 years). Short-term survival following heart transplantation did not significantly differ between women and men as 1-year survival was 89.8% (95% CI 89.1 to 90.5%) compared with 90.1% (95% CI 89.7 to 90.4%, p=0.44) (Figure 2). Likewise, 1 year CS was similar at 1, 3 and 5 years post-transplant (p=0.29, 0.75 and 0.99, respectively). Three-year CS remained similar among men and women until 10 years post-transplant, when it was 85.9% (95% CI 84.5 to 87.2%) for women as compared to 80.3% (95% CI 79.4 to 81.2%) for men (p=<0.001). Five-year CS diverged at five years post-transplant, when it was 80.4% (95% CI 79.5–81.4%) for women as compared to 77.5% (95% CI 76.9 to 78.1%) for men (p=<0.01). Women had significantly improved long-term CS at all-time points, with the absolute difference ranging from 1.8% (95% CI 1.2 to 2.5%) at the time of transplantation to 11.5% (95% CI 9.9 to 12.9%) at 5 years post-transplant. This difference persisted when the analysis was as stratified donor-recipient sex mismatch (Supplementary Figure 1).

Figure 2.

Conditional survival following heart transplantation by sex

Conditional Survival between Age Groups

Recipient age was associated with significant differences in post-transplant survival. One-year survival decreased from 91.3% (95% CI 90.8 to 91.9%) in patients aged 18–49 years to 89.9% (95% CI 89.4 to 90.3%) in patients aged 50–64 years to 87.3% (95% CI 86.3 to 88.3%) in those 65 years and older (p<0.001). Short-term CS (1-year) declined for the oldest cohort relative to the youngest cohort starting at 5 years post-transplant (absolute decrease 2.0% [95% CI 1.7 to 2.2%], p<0.001). At ten years post-transplant, 1-year CS was lower for both the 50–64 year group and the 65+ year group than the 18–49 year group (p<0.001 for both). Mid-range CS began to decline for the oldest cohort starting at 3 years post-transplant, and 5-year CS was 4.9% lower (95% CI 3.8 to 6.1%) than the 50–64y group at 5 years post-transplant. Long-term CS probabilities were significantly lower by the first year post-transplant for the two older groups compared to the youngest group (5.9% p<0.001 and 11.3% p<0.001 respectively) and the difference increased at each additional time point (Figure 3).

Figure 3.

Conditional survival following heart transplantation by age groups (18–49 years old, 50–64 years old, 65+ years old)

Conditional Survival by Mechanical Circulatory Support (MCS) Utilization

Nearly half of patients prior to transplant were bridged to transplant with MCS (percutaneous and durable) support. Patients transplanted from ECMO are known to have a significantly lower 1-year survival, and in this study had a 70.5% (95% CI 64.1 to 76.5%) survival probability compared with 90.1% (95% CI 89.8 to 90.5%) for non-ECMO patients (p<0.001). This decreased survival, however, did not persist in CS analysis. Patients transplanted from ECMO had equivalent or superior CS probabilities from one-year post-transplant onward (Figure 4A). Use of an IABP yielded similar survival post-transplant as patients who were not supported by an IABP (Supplementary Figure 2). Post-transplant survival differed significantly by type of durable MCS prior to transplant. Patients with a pre-transplant left ventricular assist device had similar short- and long-term probability of survival at time of transplant as well as in the CS analysis (Supplementary Figure 3). Patients supported with a surgical biventricular assist device (BiVAD) had decreased short-term and long-term survival at the time of transplant that ranged between 5.9 to 8.8% less (p<0.001). However, BiVAD patients had similar CS from one-year post-transplant onward. Patients supported with a total artificial heart (TAH) similarly had a decreased survival probability at the time of transplant (Figure 4C), but had decreased 1 year (3.6% absolute decrease, 95% CI 2.7 to 4.6% decrease), 3 year (3.8% absolute decrease, 95% CI 2.1 to 5.6% decrease), and 5 year CS (6.4% absolute decrease, 95% CI 3.2 to 9.8% decrease) at one-year post-transplant. This trend continued when CS was assessed at later time points but was not statistically significant due to a smaller number of patients (Figure 4C).

Figure 4.

Conditional survival by various forms of mechanical circulatory support (A) ECMO vs. no ECMO (B) Biventricular assist device vs. no mechanical circulatory support (C) Total artificial heart vs. no mechanical circulatory support

Conditional Survival with First Year Post-transplant Complications

Rejection events

Rejection during the index transplant hospitalization was categorized as none, rejection treated with additional anti-rejection agent, and rejection not treated with additional anti-rejection agent. There was no significant difference in CS at 1, 3, 5, or 10 years post-transplant between patients without rejection and those with rejection that did not require an additional agent (Supplementary Figure 4). Patients that required treatment with an additional agent, however, had a consistent and significantly decreased CS when compared with either group through the first five years post-transplant. The difference in CS only dissipated ten years post-transplant.

Rejection during the first year similarly decreased long-term survival. When limiting the analysis to only patients who were alive one-year post-transplant, patients with rejection during the first year uniformly had decreased CS (Supplementary Figure 5). One year CS (Absolute decrease of 1.8 to 5.0%, 95% CI 1.4 to 6.5% decrease) never equalized, whereas the statistically significantly reduced three year CS persisted until year 11 (Absolute decrease 2.9%, 95% CI −7.0 to 0.8%).

Major infection

Infection (of any etiology) requiring hospitalization during the first year was associated with decreased survival following transplantation. Limiting the analysis to patients alive at one year post-transplant, short (−1.4 to −1.7%, 95% CI −1.0 to −2.0%), mid (−1.9 to −6.2%, 95% CI −1.3 to −7.5%), and long term (−7.8 to −10.8%, 95% CI −6.2 to −13.7%) survival was reduced among patients who required hospitalization for infection during the first year (Supplementary Figure 6). The only exception being one year CS at year six post-transplant (0.3% absolute difference, 95% CI 0.1 to 0.4%).

Discussion:

In this study, we performed CS analysis of the UNOS registry from 2004 to 2018. CS analyses were performed utilizing time post-transplant, baseline characteristics, and post-transplant events and the main findings are as follows: (1) Among patients that survived 1 year post-transplant, 1 year CS probability is excellent and stable over the next 5 years. (2) At each time point following heart transplantation, the 3-year CS probability exceeds the 3-year actuarial survival probability. (3) Women had significantly improved post-transplant survival and higher long-term CS at each time point. (4) Recipient age was associated with significant differences in post-transplant survival. Short-term and long-term CS declined for the oldest cohort relative to the youngest cohort, with the difference increasing at each additional time point. (5) Patients supported with ECMO had significantly lower 1-year survival; however, those patients had equivalent or superior CS probabilities from one-year post-transplant onward. (6) Patients bridged with BiVAD and TAH had a decreased survival probability at the time of transplant; however, BiVAD patients had similar CS from one year post-transplant onward while TAH patients had a persistently decreased CS probability. (7) Rejection during the index transplant hospitalization that required additional immunosuppressive therapy was associated with a persistently decreased CS. (8) Infection requiring hospitalization during the first year was associated with a decreased CS probability that persisted over time.

Heart transplantation is a lifesaving procedure that affords both an excellent prognosis and quality of life. However, heart transplantation does not cure a fatal disease; rather it replaces it with a chronic disease that requires lifelong medications, careful management, and regular follow-up. This long-term relationship with the medical team frequently involves discussions of longevity and as clinicians, our answer is guided by survival curves generated from the time of the transplant. Herein, we provide a useful analysis to help guide the post-transplant discussion. The outcome in this study is conditional on both the time that has passed since transplantation, risk factors preceding the transplant, and complications that occur immediately post-transplantation. These data provide the patient and physician a more accurate expectation, specifically accounting for time elapsed post-transplant and prior transplant complications. For instance, at the time of transplant a patient’s probability to survive 1 year is 90%; however, if they have already survived 1 year their chance to survive another year is 96%, not 82% as suggested by actuarial survival curves.⁴

CS also serves a tool to differentiate between conditions that portend increased risk that is limited to the immediate post-operative period and those that carry an ongoing risk, irrespective of the time since transplant surgery. As previously reported, women have greater long-term survival than men; however, those differences are not present in the immediate post-transplant period or even during the first 5 years. The late CS benefit reported here raises questions, especially since the long-term survival difference was not explained by donor-recipient sex mismatch: are women as a whole more adherent with medical therapy and as such will benefit from better outcome in the end? Alternatively, perhaps there are genetic differences that confer a long-term survival advantage to women following heart transplantation? This finding is also intriguing, as it is known that women are at a greater risk for rejection in general,^7,8 and specifically antibody mediated rejection. The CS analysis on patients with rejection in the first year found a persistent, increased short and long-term risk associated with early rejection. Thus, further work is needed to better understand the underlying pathophysiologic differences between women and men that contribute to this sustained survival benefit over time.

Contrary to the sustained impact of sex differences on post-transplant survival, CS analysis revealed that the risk associated with ECMO or BiVAD support as a bridge to transplant is limited to the initial post-operative period. The data demonstrated that patients who survived the first year did not experience decreased CS in future years. This finding is reassuring; specifically because both ECMO and BiVAD use are associated with increased blood product utilization and infection in the pre-transplant period, raising concern that patients may pay a price not only in the immediate post-operative period, but also in the future from an increased risk for rejection due to immune activation. Furthermore, these findings support the new allocation system assigned to those supported with these devices as Status 1 designation, with the expectation that the patients who survive the early increased hazard will have a long-term prognosis equivalent to patients who were not bridged using ECMO or BiVAD. In contrast to ECMO and BiVAD, TAH was associated with an early and sustained reduction in post-transplant CS. While the number of patients supported with TAH were relatively small (n=345) and outcomes may be heterogeneous due to differing center experience and expertise,⁹ the reason for the persistent risk associated with TAH use warrants further exploration.

The early post-transplant period is known to be among the most challenging, balancing the increased risk of both rejection and infection. This analysis confirms the importance and perils of this period. Patients who developed rejection requiring treatment with an additional anti-rejection agent during the transplant hospitalization experienced not only an initial increased risk of death, but this risk persisted until 10 years post-transplant. In the same way, patients who experienced rejection requiring hospitalization during the first year post-transplant had decreased CS, even when excluding the analysis to only patients who survived the first year. The delicate balance extended to infection where an infection requiring hospitalization during the first year carried an increased risk of mortality extending to ten years post-transplant. These results reinforce the long-term impact of both rejection and infection events in the first year post-transplant.

This study is not without limitations. First, while the UNOS dataset used for this analysis is high quality in that for all U.S. transplant centers data submission is mandatory by law, the data is subject to reporting inaccuracies and is limited to the data available for analysis. Next, conditional survival analysis offers a different way to analyze and present survival data, however it is inherently limited to analysis for a single risk factor. We acknowledge that limiting analysis to a single risk factor does not account for the potential of multiple risk factors contributing to long-term survival. Further, the 14-year study period was selected to allow for 10-year conditional survival analysis; however, this generated a study period where the median survival has increased from 12.5 to a projected 14 years. Lastly, the analysis identified two events (rejection and infection) that are captured by the UNOS registry from the early post-transplant period that contribute significantly to post-transplant morbidity and mortality, however additional events that can impact long-term survival (e.g. CAV and malignancy) were not assessed.

In conclusion, the use of CS analysis has provided a tool to enhance patient and provider discussions regarding survival expectations post primary heart transplant. This analysis highlights the importance of age, sex, and perioperative and short-term events on long-term outcome. Furthermore, this study have shown that actuarial survival curves that are currently used to predict longevity post-heart transplant can be improved in an effort to better inform our patients and health care team.

Supplementary Material

Supplemental Material

Supplementary Figure 1. Conditional survival following heart transplantation stratified by donor-recipient sex pairing.

Supplementary Figure 2. Conditional survival following heart transplantation stratified by intraaortic balloon pump use.

Supplementary Figure 3. Conditional survival following heart transplantation stratified by durable LVAD use.

Supplementary Figure 4. Conditional survival following heart transplantation stratified by rejection during the transplant hospitalization.

Supplementary Figure 5. Conditional survival following heart transplantation of patients who survived at least one year, stratified by rejection during the first year post-transplant.

Supplementary Figure 6. Conditional survival following heart transplantation of patients who survived at least one year, stratified by hospitalization for infection during the first year post-transplant.